Quantification of Joint Blood Flow by Dynamic

Western University Scholarship@Western

Physiology and Pharmacology Publications Physiology and Pharmacology Department

1-1-2020

Quantification of joint blood flowy b dynamic contrast-enhanced near-infrared spectroscopy: Application to monitoring disease activity in a rat model of rheumatoid arthritis

Seva Ioussoufovitch Western University

Laura B. Morrison Lawson Health Research Institute

Lise Desjardins Lawson Health Research Institute

Jennifer A. Hadway Lawson Health Research Institute

Keith St Lawrence Lawson Health Research Institute

See next page for additional authors

Follow this and additional works at: https://ir.lib.uwo.ca/physpharmpub

Part of the Medical Physiology Commons, and the Pharmacy and Pharmaceutical Sciences Commons

Citation of this paper: Ioussoufovitch, Seva; Morrison, Laura B.; Desjardins, Lise; Hadway, Jennifer A.; Lawrence, Keith St; Lee, Ting Yim; Beier, Frank; and Diop, Mamadou, "Quantification of joint blood flow by dynamic contrast- enhanced near-infrared spectroscopy: Application to monitoring disease activity in a rat model of rheumatoid arthritis" (2020). Physiology and Pharmacology Publications. 117. https://ir.lib.uwo.ca/physpharmpub/117 Authors Seva Ioussoufovitch, Laura B. Morrison, Lise Desjardins, Jennifer A. Hadway, Keith St Lawrence, Ting Yim Lee, Frank Beier, and Mamadou Diop

This article is available at Scholarship@Western: https://ir.lib.uwo.ca/physpharmpub/117 Quantification of joint blood flow by dynamic contrast-enhanced near- infrared spectroscopy: application to monitoring disease activity in a rat model of rheumatoid arthritis

Seva Ioussoufovitch Laura B. Morrison Lise Desjardins Jennifer A. Hadway Keith St. Lawrence Ting-Yim Lee Frank Beier Mamadou Diop

Seva Ioussoufovitch, Laura B. Morrison, Lise Desjardins, Jennifer A. Hadway, Keith St. Lawrence, Ting- Yim Lee, Frank Beier, Mamadou Diop, “Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy: application to monitoring disease activity in a rat model of rheumatoid arthritis,” J. Biomed. Opt. 25(1), 015003 (2020), doi: 10.1117/1.JBO.25.1.015003.

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Journal of Biomedical Optics 25(1), 015003 (January 2020)

Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy: application to monitoring disease activity in a rat model of rheumatoid arthritis

Seva Ioussoufovitch,a Laura B. Morrison,b Lise Desjardins,b Jennifer A. Hadway,b Keith St. Lawrence,b,c Ting-Yim Lee,b,c,d Frank Beier,e and Mamadou Diopa,b,c,* aWestern University, Bone and Joint Institute, School of Biomedical Engineering, Faculty of Engineering, London, Ontario, Canada bLawson Health Research Institute, Imaging Program, London, Ontario, Canada cWestern University, Schulich School of Medicine and Dentistry, Department of Medical Biophysics, London, Ontario, Canada dRobarts Research Institute, Imaging Program, London, Ontario, Canada eWestern University, Schulich School of Medicine and Dentistry, Department of Physiology and Pharmacology, London, Ontario, Canada

Abstract Significance: Current guidelines for rheumatoid arthritis (RA) management recommend early treatment with disease modifying antirheumatic drugs (DMARDs). However, DMARD treatment fails in 30% of patients and current monitoring methods can only detect failure after 3 to 6 months of therapy. Aim: We investigated whether joint blood flow (BF), quantified using dynamic contrast-enhanced time-resolved near-infrared spectroscopy, can monitor disease activity and treatment response in a rat model of RA. Approach: Ankle joint BF was measured every 5 days in eight rats with adjuvant-induced arthritis (AIA) and four healthy controls. Arthritis was allowed to progress for 20 days before rats with AIA were treated with a DMARD once every 5 days until day 40. Results: Time and group had separate significant main effects on joint BF; however, there was no significant interaction between time and group despite a notable difference in average joint BF on day 5. Comparison of individual blood flow measures between rats with AIA and control group animals did not reveal a clear response to treatment. Conclusions: Joint BF time courses could not distinguish between rats with AIA and study controls. Heterogeneous disease response and low temporal frequency of BF measurements may have been important study limitations. © The Authors. Published by SPIE under a Creative Commons Attribution 4.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI. [DOI: 10.1117/1.JBO.25.1.015003] Keywords: time-resolved near-infrared spectroscopy; dynamic contrast-enhanced; blood flow; rheumatoid arthritis; treatment monitoring; disease-modifying antirheumatic drugs. Paper 190146RR received May 8, 2019; accepted for publication Dec. 6, 2019; published online Jan. 14, 2020.

1 Introduction Treatment response is currently assessed using a combina- Rheumatoid arthritis (RA) is a chronic and progressive auto- tion of clinical examination and patient self-assessment. Though this approach is the current standard of care, it can be ineffective immune disease that afflicts about 1% of the population and is for repetitive, longitudinal assessments because of its subjec- associated with pain,1 reduced quality of life,2 and disability.3 tivity and variability.9,10 Furthermore, clinical examination and Fortunately, the introduction of a treatment paradigm that patient self-assessment are unlikely to detect subclinical combines treat-to-target strategies4 with early use of disease- changes in inflammation, which could be early indications of modifying antirheumatic drugs (DMARDs)5 has significantly treatment response. As such, clinical examination is often sup- improved mid- and long-term patient outcomes over the past plemented with laboratory tests, radiography, or ultrasonogra- two decades. However, DMARD treatment failure, which is cur- phy. However, though these additional tools can be effective rently identified only after 3 to 6 months of therapy, still occurs at aiding diagnosis, their usefulness for long-term monitoring in 30% of RA patients.6,7 After failure, these patients must is limited by low sensitivity (radiography and laboratory undergo an iterative process where they are assigned to new – tests)11 and suboptimal reproducibility (ultrasonography).12 14 therapies and wait again for 3 to 6 months before treatment effi- Given the above limitations, there is currently a need to find cacy can be reliably assessed. This process is drawn-out, costly, more sensitive and objective techniques that can detect RA treat- and results in patients losing the benefits of effective early treatment failure within the first 3 months of therapy. ment while having an increased risk of developing irreversible 8 In recent years, there has been a growing interest in inves- joint damage. tigating the ability of near-infrared (NIR) optical methods to assess RA disease progression. This is in part due to the ability of NIR techniques to provide quick and objective measurements *Address all correspondence to Mamadou Diop, E-mail: [email protected] at a relatively low cost. In the case of RA, NIR methods

Journal of Biomedical Optics 015003-1 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

generally attempt to monitor disease progression by measuring functions (IRFs) were acquired by placing the emission and downstream effects of joint hypoxia, which is well known to detection probes into a light-tight box containing a piece of play a central role in the maintenance and progression of the paper, positioned 2 mm away from the emission fiber, which disease.15,16 It is well established that chronic inflammation, acted as a light diffuser to fill the numerical aperture of the which is a key feature of RA, causes hypertrophy of the synovial detection probe. lining; the thickness of the synovial membrane increases from 16 1 to 3 to over a dozen cell layers in RA. This hypertrophic cell 2.2 Phantom Experiments mass induces an increased metabolic demand that exceeds sup- ply, leading to the development of hypoxic regions within the Since the DCE TR-NIRS technique relies on accurate estimation synovium.15,17 In fact, it has been shown (using invasive probes) of changes in tissue absorption, we assessed the system’s that the partial pressure of oxygen is lower in inflamed joints ability to measure absorption changes in a tissue-mimicking than in healthy joints.18 The presence of chronic hypoxia then phantom made with Intralipid and India Ink. A cuvette acts as a signal both for the formation of new blood vessels (10 mm × 10 mm × 40 mm) was filled with 3.5 mL of 0.8% (i.e., angiogenesis)17 and increased tissue blood flow (BF).19,20 Intralipid created by diluting a 20% stock solution of Intralipid Previous studies using NIR techniques have largely focused with water. We then placed the emission and detection probes on assessing changes in oxygen saturation and blood content transversely across the cuvette, secured the entire setup using a within the joint since these characterize hypoxia and angiogen- 3-D-printed holder, and acquired a baseline measurement with esis, respectively, and are the downstream effects of chronic the TR-NIRS system. Next, we began adding 0.02-mL incre- inflammation. Some examples of NIR techniques that have been ments of an India Ink solution with a known absorption coef- μ μ applied to RA disease monitoring include photoacoustic tomog- ficient ( a;ink). Note that a;ink was determined by measuring raphy,21–24 diffuse optical spectroscopy,25–27 contrast-based fluo- light transmittance through an India ink dilution to determine rescence imaging,28–30 and diffuse optical tomography.31–34 the ink’s molar absorption coefficient (see Sec. 2.4 for details). Using these approaches, RA disease progression can be assessed The volume of increments was chosen so that the investigated by measuring increases in blood volume, variations in blood changes in absorption would cover a range similar to what had oxygen saturation, intensity of contrast agent fluorescence, and been measured during preliminary animal experiments (0.005 to total hemoglobin concentration in the joint. As mentioned, 0.040 cm−1). After adding each 0.02-mL increment, the solu- increased tissue BF—defined as the volume of blood flowing tion was mixed with a glass stirring rod and left to settle for through a mass of tissue per unit time (mL/min/100 g)—is 60 s before the next set of measurements. Each set of TR- another physiological response to chronic hypoxia; in fact, joint NIRS measurements consisted of 100 distributions of time- BF has yet to be thoroughly investigated as a potentially highly of-flight (DTOFs) acquired over 30 s. sensitive marker of RA disease activity. Thus, we hypothesized that BF can be a surrogate marker of changes in joint hypoxia 2.3 Animal Model and Experiments and RA disease progression. We previously developed a dynamic contrast-enhanced time-resolved near-infrared spec- Animal experiments were conducted under an animal care pro- troscopy (DCE TR-NIRS) technique for measuring joint BF and tocol approved by the local ethics committee. All experimental showed—in a rabbit model of RA—that joint BF is more sen- procedures were conducted while the animals were under anes- sitive to inflammatory arthritis than changes in hemoglobin con- thetic. Anesthesia was induced in an airtight chamber with 5% centration and oxygen saturation.35 These findings suggest that isoflurane gas and maintained with 2% isoflurane by mask. joint BF may be a more sensitive biomarker of inflammatory The study included 12 adult male Lewis rats: four rats in the arthropathies than those previously investigated with other control group and eight rats in the experimental group (exper- NIR techniques. Thus, the objective of this work was to inves- imental timeline is shown in Fig. 1). Arthritis was induced using tigate whether joint BF, as measured with DCE TR-NIRS, can the adjuvant-induced arthritis (AIA) model.36 Prior to the start of track longitudinal changes in joint inflammation during disease the study, animals were allowed to acclimate for 5 to 7 days, induction and DMARD treatment in a rat model of RA. after which three baseline BF measurements were acquired over a 10-day period (Fig. 1). After the third baseline measurement 2 Methods was acquired on day 0, rats in the experimental group received a subcutaneous injection of a solution prepared using heat-killed 2.1 Instrumentation lyophilized Mycobacteria butyricum suspended in incomplete Freund’s adjuvant to induce polyarticular arthritis;37 rats in the The TR-NIRS system was built in-house using a pulsed diode control group received saline injections. Thereafter, ankle joint laser (LDH-P-C-810; PicoQuant, Germany) connected to a PDL BF was measured every 5 days until the end of the study on day 828 laser driver (PicoQuant). The laser emission was centered at 40 or until rats reached predetermined humane endpoints (HEP) 805 nm and the pulse repetition rate was set to 80 MHz. The based on pain assessment by a veterinarian. AIA was allowed to laser output was coupled into an emission fiber (ϕ ¼ 400 μm, progress until day 20 (pretreatment phase). Starting on day 20, NA ¼ 0.22; Fiberoptics Technology, Pomfret, Connecticut) that the treatment phase of the study began: rats that had not reached guided light to the ankle joint. Light transmitted through the HEP were treated with intramuscular injections of the DMARD ankle joint was collected with a fiber optic bundle (ϕ ¼ 3mm, etanercept (Enbrel®: 0.5 mL/kg) every 5 days. Treatments were NA ¼ 0.55; Fiberoptics Technology) that was coupled to a administered on each measurement day within the treatment hybrid photomultiplier detector (PMA Hybrid; PicoQuant) and phase immediately after BF measurements. Rats in the experi- a time-correlated single photon-counting module (HydraHarp mental group were studied in separate cohorts of 2; the entire 400; PicoQuant). A three-dimensional (3-D)-printed probe- study protocol (e.g., induction, treatment) was repeated four holder was used to ensure good contact between the ankle joint times with cohorts of 2 rats for a final sample size of 8 in the and the emission and detection probes. Instrument response experimental group. Rats in the control group were studied

Journal of Biomedical Optics 015003-2 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

Fig. 1 Experimental timeline. Baseline measurements (B1 to B3) were acquired within a 10-day period before and on the day of arthritis induction in the experimental group (day 0). Starting on day 20, animals in the experimental group were treated with the DMARD Enbrel® (etanercept) every 5 days.

simultaneously in one cohort of four animals. Weight measure- the tissue ICG concentration curve. On every measurement ments and qualitative written notes describing animal appear- day, joint BF was measured on both sides of the animal (i.e., ance, behavior, gait, and swelling in the paws or joints were right and left ankles); measurements were repeated twice to recorded for each rat on and between study days by two certified mitigate potential unsuccessful measurements. Typical reasons veterinary technicians (L. Morrison and L. Desjardins). These for unsuccessful measurements included data loss due to DDG observations were used in consultation with a veterinarian to instrument failure, failed ICG injections due to poor catheter determine the appearance of first symptoms of induced inflam- placement, and loss of probe contact (both DDG and TR-NIRS) mation and to assess whether an animal had reached HEP. during acquisition. Evidence of arthritis included visible swelling in paws or joints, weight loss (>5% of initial body weight), reluctance to ambu- 2.4 Data Analysis late, decreased social interaction, and abnormal posture. The experimental setup for the BF measurements is shown in The TR-NIRS measurements were analyzed using an in-house software developed in MATLAB 2017a (The MathWorks Inc., Fig. 2. During each measurement, animals received a tail vein Natick, Massachusetts, 2017). For each measurement, the differ- bolus injection of the optical contrast agent Indocyanine Green ence in mean photon time-of-flight (hti) between the DTOFs (ICG). A dye densitometer (DDG-2001; Nihon Kohden, Japan) acquired prior to contrast introduction (India Ink for phantom was attached to one paw to measure the arterial concentration experiments; ICG injection for animal experiments) and the sys- of ICG. The dye densitometer also measured the animal’s heart tem’s IRF was used to compute the optical pathlength p rate (HR) and arterial oxygen saturation, which were recorded c and used to assess animal condition throughout the experiment. p ¼ ðhti − hti Þ: TR-NIRS measurements were acquired with the emission and EQ-TARGET;temp:intralink-;e001;326;408 n DTOF IRF (1) detection probes positioned transversely across the rat ankle joint on the contralateral paw (Fig. 2). Each joint BF measure- In Eq. (1), c is the speed of light in vacuum, and n is the ment was started by acquiring 10 s of data before a bolus of ICG refractive index of tissue (n ¼ 1.4). The optical pathlength was solution (0.2 mg/kg) was injected into the rat tail vein. For each then used, in combination with the modified Beer–Lambert BF measurement, we acquired 400 DTOFs over 120 s to obtain Law, to determine changes in the absorption coefficient over time [i.e., ΔμaðtÞ]: IðtÞ Δμ ðtÞ¼ ÷ p: EQ-TARGET;temp:intralink-;e002;326;313 a ln I (2) 0 IðtÞ I In Eq. (2), and 0 are the detected light intensities after and prior to introduction of the contrast agent, respectively. Note that light intensities were computed as the sum of the total number of photons in each DTOF and that, for animal experiments, changes in pathlength due to ICG injection were deemed negligible as discussed in Sec. 4. For phantom experiments, the measured Δμa were compared with expected Δμa for validation. The expected Δμa were computed using Eq. (3). The molar ε absorption coefficient of the ink ( a;ink) was calculated by deter- ε mining the molar extinction coefficient e;ink from a transmission measurement through India ink of known dilution and ε multiplying it by the ratio a;ink ¼ 0.88538 as suggested by εe;ink Spinelli et al:39 Δμ ¼ ð Þε Δc; ε ¼ ε : EQ-TARGET;temp:intralink-;e003;326;126 a;expected ln 10 a;ink a;ink 0.885 e;ink (3) Fig. 2 Image of TR-NIRS probe placement on a rat ankle for joint BF measurement. A dye densitometer, placed on the contralateral paw, was used to measure the time-dependent arterial concentration of the For joint measurements, it was noted that the geometry of BF tracer (ICG) while TR-NIRS probes were used to measure its tis- the joint, along with its expected structural changes during dis- sue concentration. ease progression, substantially complicates the use of typical

Journal of Biomedical Optics 015003-3 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

analytical solutions of light propagation in tissue and fitting was then used to assess the power of the findings using an effect approaches to determine optical properties (see Sec. 4 for size equal to the partial η2 value determined from the ANOVA. detailed discussion). Instead, use of the modified Beer–Lambert The potentially confounding effects of HR on joint BF were Law to compute changes in absorption provides a robust method investigated using correlation analysis. Since joint BF could be that is immune to the challenges posed by the joint morphology. affected by arthritis induction and DMARD treatment, as well as Furthermore, the measured ΔμaðtÞ was converted into an ICG other temporal factors, the effect of HR on joint BF could only tissue concentration curve using be isolated on a per day basis. Thus, for timepoints with no missing data, each animal’s raw daily BF measurements were corre- Δμ ðtÞ QðtÞ¼ a ; lated with their corresponding HR to generate a correlation EQ-TARGET;temp:intralink-;e004;63;675 ð Þ ε (4) ln 10 · ICG coefficient between BF and HR for each day. Since each correlation coefficient was derived using only four comparisons (one where QðtÞ is the time-dependent ICG concentration in the for each raw measurement), the resulting coefficients were ε ankle joint, and ICG is the extinction coefficient of ICG at prone to error and difficult to interpret in isolation; instead, all 805 nm.40 The measured tissue and arterial ICG concentration obtained correlation coefficients were averaged with the curves were subsequently used to compute joint BF, using a pre- assumption that the presence of an underlying trend in the data viously reported deconvolution algorithm.41 Note that this would skew the average correlation coefficient. method of computing BF has been previously tested and vali- dated using phantom and animal experiments.42,43 As mentioned 3 Results in Sec. 2.3, four perfusion measurements (two per ankle joint) were obtained for each animal in order to account for potentially 3.1 Phantom Experiments unsuccessful data collection. If both repeat measurements from A simple linear regression with the y-intercept set to 0 (see one ankle were available, the BF value for that ankle on that day Sec. 2.5 for details) was conducted to compare Δμa measured was calculated as the mean of the two measurements. It is note- worthy that in ∼10% of cases, one out of the two repeat measurements was missing and only the nonmissing value was used.

2.5 Statistical Analysis Statistical analysis was conducted using SPSS Statistics 25 (IBM Corp., Armonk, New York, 2017), and power analysis was performed using G*Power software.44 For the phantom experiments, agreement between expected and measured changes in μa was investigated using linear regression. Regression analysis was performed using a y-intercept fixed Δμ ¼ Δμ ¼ at 0 with the assumption that a;observed a;expected 0 prior to the addition of any India ink. Prior to analysis, regression assumptions of linearity and homoscedasticity were confirmed. For the linearity assumption, a visual inspection of scatter plots of raw observed versus expected values was used Fig. 3 Comparison of expected and measured changes (mean SD) to confirm a linear relationship between the two variables. Data in the absorption coefficient of a tissue-mimicking solution (0.8% homoscedasticity was confirmed by generating scatter plots of Intralipid) caused by incremental addition of India Ink. One hundred residuals versus values predicted by the regression model and measurements were acquired using the time-resolved NIR spectros- visual confirmation of consistent variance of the residuals for copy system, and absorption coefficients were computed using the – all predicted values. Furthermore, assumption of normality in modified Beer Lambert Law. Note that absorption changes measured during subsequent animal experiments were within this range. the dataset was confirmed using the Shapiro–Wilk test. For the in vivo experiments, statistical analysis was only conducted for timepoints that contained data from all animals, to account for animals reaching HEP before study completion. Though this approach limited the scope of the statistical analysis, it was necessary in order to avoid any potential survivorship bias within the dataset. A three-way repeated measures analysis of variance (ANOVA) was conducted with time and measurement side (i.e., left or right) as the within-subjects variables, and subject group as the between-subjects variable (i.e., control or experimental group). Prior to analysis, normality of the dependent variable and sphericity were confirmed, using the Shapiro– Wilk test and Mauchly’s test of sphericity, respectively, to ensure no assumptions inherent to ANOVA were violated. In addition, visual inspection of boxplots and histograms of BF data within each group revealed no significant outliers. Upon discovering a significant effect, differences were uncovered Fig. 4 Arterial and tissue ICG concentration curves measured by the using a post-hoc Tukey’s honest significant difference test to dye densitometer and the TR-NIRS system, respectively. ICG was account for multiple comparisons. A post-hoc test in G*Power injected into the rat’s tail vein at the 10-s mark.

Journal of Biomedical Optics 015003-4 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

in tissue-mimicking solution with the expected Δμa due to India 3.2 Animal Experiments ink addition. There was a significant linear relationship between Figure 4 shows typical arterial and tissue ICG concentration measured and expected Δμa [Fð1;9Þ¼5773, p < 0.05], and a strong relationship between the variables (R2 ¼ 0.99) with a curves measured during an animal experiment. The dynamic slope of 0.96 within the tested 0.005 to 0.040 cm−1 absorption inflow and washout of the tracer can be clearly identified on range (Fig. 3). both curves, with the tissue curve showing a slower rise and subsequent clearing of ICG compared with the arterial curve. The results of the statistical analysis from the first day of the Table 1 Summary of three-way repeated measures ANOVA with baseline period (B1) to day 15 are summarized in Table 1. In all time and measurement side as the within-subject variables, and group cases, the BF data were not significantly different from a normal as the between-subjects variable. Analysis was conducted on data distribution (p ¼ ns; ns, not significant) and did not violate the from the first day of the baseline period (B1) to day 15 postinduction assumption of sphericity (p ¼ ns). Only time and group were of arthritis. found to have a significant main effect on joint BF (Table 1) with effect sizes of partial η2 ¼ :400 and η2 ¼ :365, respec- Variable(s)a F-statistic p-Valueb tively. These effects were interpreted as follows: BF increased over time regardless of group assignment, and experimental timeagroupaside Fð5;50Þ¼0.149 p ¼ ns group BF was higher than control group BF regardless of measurement time. Figures 5(a) and 5(b) show the data relevant to the timeaside Fð5;50Þ¼0.659 p ¼ ns statistically significant effects denoted in Table 1: a BF time timeagroup Fð5;50Þ¼1.773 p ¼ ns course averaged over measurement side and group, and group BF values averaged over time and measurement side. As seen in a group side Fð1;10Þ¼0.135 p ¼ ns Fig. 5(a), a post-hoc Tukey test for the effect of time on ankle Time Fð5;50Þ¼6.670 p < 0.05 joint BF revealed a significant difference between BF on the first day of baseline (B1) and BF on day 10 at the p < 0.05 level. In Group Fð1;10Þ¼5.736 p < 0.05 addition, since statistical analysis revealed that measurement side was not implicated in any interaction and did not have a Fð1;10Þ¼0 129 p ¼ side . ns significant main effect on BF (Table 1), time courses for each aInteraction between multiple variables group averaged over measurement side are presented in Fig. 5(c) bns: not significant to facilitate interpretation of the uncovered main effects. We

Fig. 5 (a) BF (95% confidence interval) for time courses averaged by group and measurement side, and (b) boxplots of BF for each group averaged over time and measurement side. Asterisks indicate a significant difference at the p < 0.05 level between (a) two timepoints or (b) two groups. To facilitate interpretation of main effects, BF (95% confidence interval) time courses for each group averaged over measurement side are shown in (c). Time courses (a) and (c) show three baseline measurements (B1 to B3; gray), followed by three measurements postarthritis induction in the experimental group (after B3: day 0; orange). Note that, to avoid survivorship bias, statistical analysis was only conducted on timepoints that included all animals in each group; thus, data from the treatment phase are not shown.

Journal of Biomedical Optics 015003-5 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

note that the primary effect of interest in this work was the pres- variation in their joint BF compared with other rats in their ence of a two-way interaction between time and group, which group, and their BF times courses are generally more similar would have manifested as a significant difference between the to those of the rats in the control group. In contrast, for rats time courses in Fig. 5(c). The presence of this effect would have 1 and 2 joint BF decreased once treatment was initiated; how- been interpreted as a difference in joint BF between the exper- ever, only a few treatment timepoints are available since these imental and control group over time. To further investigate the rats reached HEP quickly. Looking at individual side measure- lack of this effect (Table 1), post-hoc power analysis was calcu- ments, there is no clear trend in the inconsistency between lated for the two-way interaction between time and group using measurement sides across animals. For the experimental an effect size of partial η2 ¼ 0.151 and yielded a power of 0.21. group, average BF time courses among rats are more similar Confounding effects of HR on measured joint BF values within each cohort than between cohorts. This similarity is also were investigated by correlating raw BF measurements with present in the occurrence of first symptoms and reaching of their corresponding HR and averaging across all animals and HEP (Fig. 6). days as described in Sec. 2.5. This analysis revealed a weak positive correlation (r ¼ 0.1370) between HR and BF 4 Discussion measurements. This work sought to investigate whether joint BF, as measured To further investigate the variations in joint BF, longitudinal with DCE TR-NIRS, could track longitudinal changes in joint measurements from each individual rat were examined (Fig. 6); inflammation during disease induction and DMARD treatment time courses for raw data as well as data averaged across meas- in the AIA rat model of RA. AIA was the first-described animal urement side are presented. Consistently elevated joint BF levels model of RA and remains widely used for preclinical assessment throughout the pretreatment phase were only observed for rats of RA treatments.45 The model is typically characterized by 1 and 2; these animals had the highest incidence of ankle BF rapid onset (e.g., 10 days postinjection) and progression of pol- above the typical baseline threshold of ∼15 mL∕ min ∕100 g. yarticular arthritis; however, the inflammation tends to subside For four out of eight rats in the experimental group, joint BF after a month of disease activity. While the AIA rat model does on day 5 was higher than their typical baseline values. In the not exhibit the chronic disease progression characteristic of control group, joint BF on day 10 for was higher than typical human RA, which takes place over months, it shares many baseline values for three out of four rats. Rats 3 and 4 were the of the relevant biological features, such as swelling, joint only individuals in the experimental group who completed the destruction, cell infiltration, and T-cell dependence. As well, entire study period; these rats also showed very minimal treatment with etanercept—a common biologic drug that is

Fig. 6 Scatter plots of individual BF values (mL/min/100 g) for each rat; average BF values are shown as black markers, whereas right and left ankle measurements are indicated using red and blue markers, respectively. Time courses show three baseline measurements (B1 to B3; gray), followed by measurements post-arthritis induction in the experimental group (after B3: day 0; orange) and after DMARD treatment was started (days 20 to 40; blue). Experimental and control group data are enclosed by red and blue borders, respectively. For the experimental group, HEP and appearance of first symptoms are indicated by solid and dashed vertical lines. Rats in the experimental group were studied in cohorts of 2 while those in the control group were studied in one cohort of 4; experimental group data are subdivided by black lines to delineate cohorts. Cohort numbers for both groups are indicated at the top right (C1 to C5).

Journal of Biomedical Optics 015003-6 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

currently used to treat RA patients—is known to be active in this both groups to be statistically different from one another (i.e., no model and has been associated with slight reductions in arthritic significant interaction between time and group). and radiographic scores.36,46 It is important to note that rats tend As mentioned in Sec. 1, it is well established that chronic to develop antibodies against etanercept; thus, when considering inflammation, which is a key feature of RA, leads to the devel- the difference in body weight, response to the drug is limited to opment of hypoxic regions within the synovium15,17 and that the higher treatment doses (e.g., 4 to 10 mg every 3 to 5 days)21,46 presence of hypoxia acts as a potent signal for angiogenesis17 than what is typically used in RA patients (50 mg weekly). and increased tissue BF.19,20 Based on this known pathophysi- For proper quantification of joint BF, DCE TR-NIRS relies ology of inflammatory arthritis and the results from our previous 35 on an accurate measurement of the dynamic μa changes in a study in a rabbit model of inflammatory arthritis, we expected tissue bed following an ICG bolus injection. These measure- to find an interaction between time and group in this study. More ments can be used to compute a tissue ICG concentration curve, specifically, we anticipated an increase in joint BF in the exper- which can then be combined with an arterial ICG concentration imental group compared with the control group during the pre- curve to calculate BF. As such, we conducted tissue-mimicking treatment phase. phantom experiments to ensure that our TR-NIRS system could One possible reason that this study did not find a significant accurately measure changes in μa. Figure 3 shows a strong linear difference between control and experimental group time courses relationship (R2 ¼ 0.99) with a slope of 0.96 between expected was the introduction of measurement error due to the variation in and measured μa changes in a tissue-mimicking phantom whose animal HR during different BF measurements. However, only a absorption coefficient was modulated using various concentra- very weak correlation (r ¼ 0.1370) was noted between HR and tions of India Ink; these results confirmed our system’s ability joint BF, suggesting that the potentially confounding effect of to quantify static changes in μa. Next, the ability of the TR- varying HR throughout an experiment had a negligible effect NIRS system to track dynamic changes in absorption was con- on the BF results. Another reason could be the relatively small firmed through preliminary animal experiments during which sample size. Based on the effect size observed here for the two- representative arterial and tissue ICG concentration curves were way interaction between time and group (partial η2 ¼ 0.151), obtained. Tissue ICG concentrations were ∼100 times lower post-hoc power analysis revealed that, in regards to this two-way than the concentrations of the tracer in arterial blood (Fig. 4), interaction, our study only had a power of 0.21. To reach a which was similar to what we previously reported for tissue con- power of 0.8 with the aforementioned effect size, our study centration values in the rabbit knee joint.35 would require a sample size of 74; this is a notable difference Following system validation, experiments were conducted in from the large effect size we previously measured in an inflam- 12 rats: four rats in the study’s control group and eight rats in the matory monoarthritis rabbit model,35 where only four subjects AIA experimental group. To account for potential injection were needed to detect a statistically significant difference effects, rats in the control group received saline injections to between control and inflamed joints at the p < 0.05 level with match any injections administered to the experimental group. a power of 0.8. In particular, this discrepancy between both stud- Due to the severity of the AIA model, 50% of rats in the exper- ies highlights the challenge of modeling human disease with ani- imental group reached HEP before day 20 postinduction of mal models. arthritis. Thus, to avoid survivorship bias, we limited the tem- Aside from differentiating between groups during the pre- poral scope of our subsequent statistical analysis to timepoints treatment phase, this study sought to investigate whether treat- that contained data from all animals, i.e., all baseline timepoints ment with etanercept affects joint BF in the AIA model and the first three timepoints of the pretreatment phase. The only postinduction of arthritis. Since etanercept is known to be active significant effects found in this study were that both time and in this model of arthritis,36,46 we expected to see decreased joint group had a significant main effect on joint BF with effect sizes BF in response to etanercept treatment. However, because no of η2 ¼ 0.400 and η2 ¼ 0.365, respectively [Figs. 5(a) and 5(b)]. statistical analysis could be performed for timepoints in the Interestingly, the implication of the main effect denoted in treatment phase, we compared treatment response between con- Fig. 5(b) is that BF differed between rats in both groups regard- trol and experimental group rats on a case-by-case basis (Fig. 6). less of when it was measured (i.e., even at baseline). To further Only rats from cohorts 1 and 2 progressed into the treatment investigate this possibility, we compared mean BF time courses phase, and only rats from cohort 1 showed evidence of BF of both groups along with their 95% confidence intervals decrease in response to treatment. In particular, their BF dropped [Fig. 5(c)]; since measurement side did not have a significant from the elevated pretreatment values on days 15 and 20 to val- effect on BF, we averaged time courses across both ankles to ues similar to cohort 2 and control rats on days 25 and 30. Rats facilitate data interpretation. First, BF appears consistent in cohort 2 did not have elevated pretreatment BF values, and between both groups and among the baseline timepoints their BF remained relatively constant throughout the study. In [Fig. 5(c)]; this is an important result as it confirms the ability fact, aside from a slight elevation in BF in rat 3 on day 5, the of our DCE TR-NIRS technique to reliably quantify joint BF. time courses of the rats in cohort 2 looked quite similar to those Second, once arthritis is induced (pretreatment), BF in the of the control group. Consultation of recorded animal well- experimental group is substantially higher than in the control being metrics revealed that rats 3 and 4 had the latest appearance group on day 5 after which BF in the control group rises to of first symptoms (Fig. 6) and, while both rats exhibited some match that of the experimental group. These trends in Fig. 5(c) paw swelling, they did not have the gait issues that were typi- are consistent with the significant effect seen in Fig. 5(a); cally seen in the other cohorts. Thus, the relatively mild disease regardless of AIA, joint BF changes as the study period pro- induction in cohort 2 rats may have played some role in making gresses. Furthermore, Fig. 5(c) suggests that the significant it difficult to distinguish between them and control group rats on effect described by Fig. 5(b) is largely driven by the discrepancy the basis of joint BF. Importantly, the approach that is typically in BF between both groups on day 5. Nevertheless, this single reported in the literature is to administer treatment on the discrepancy was not substantial enough for the time courses in first day of inflammatory symptoms. In contrast, our study was

Journal of Biomedical Optics 015003-7 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

designed so that treatment was only administered at a set time- tissue perfusion;48–50 for example, DCS generally provides a point regardless of first symptom appearance. Since some vari- BF index whose proportionality to absolute BF is a function of ability in disease progression is expected and was observed tissue optical properties and tissue geometry. Importantly, RA between rats, it is possible that experimental group rats received progression is associated with synovial lining hypertrophy, treatment while in a different state of disease progression, which typically leading to bone resorption, membrane inflammation, may have confounded the treatment phase results. Due to these edema, and other general morphological changes to the joint factors as well as the small experimental group sample size tissue, which makes modeling the joint geometry quite available from the treatment phase, the conclusions regarding challenging.45 Considering that these changes vary from subject the relationship between joint BF and DMARD treatment to subject, accurate quantification of tissue optical properties response remain limited. during disease progression and treatment with DCS could be We also examined the pretreatment phase of Fig. 6 for any very difficult. Furthermore, failing to account for these changes consistent differences between the experimental and control would likely confound the relationship between BF indices mea- groups. Four out of eight rats in the experimental group showed sured at different timepoints. There are systems that combine BF increases on day 5, whereas increases from baseline only DCS with TR-NIRS or frequency-domain NIRS to make it pos- occurred on day 10 in three out of four rats in the control group; sible to quantify the necessary optical properties for continuous this is consistent with the BF difference on day 5, as shown in DCS measurement on a day-by-day basis;48,51,52 however, these Fig. 5(c). A striking observation from Fig. 6 was that, for the approaches still typically rely on the use of analytical solutions entire study period, time courses among rats within the same to the diffusion equation for simple tissue geometries (e.g., slab cohort were more similar than rats from different cohorts. geometry). As mentioned in Sec. 2.4, a key advantage of our Figure 6 also revealed that rats within the same cohort experi- DCE TR-NIRS technique is that it only requires the ability enced first symptoms within a day of each other while first to quantify changes in absorption to measure joint BF. This cir- symptom occurrence had a standard deviation of 2.20 days cumvents the challenge of needing an analytical solution to the when all cohorts were combined. Together, these observations diffusion equation for a joint structure with a complex shape that suggest that rats in the same cohort had similar disease progres- changes both in time and across subjects and is not well approxi- sion and that the heterogeneity in the response to arthritis induc- mated by regular geometrical shapes. tion between cohorts may have had an important influence on While DCE TR-NIRS has unique advantages that make it the BF time courses. This observation is further confirmed by suitable for longitudinal joint BF monitoring, a potential source the temporal consistency between first symptom appearance and of error while using this technique is the occurrence of dynamic HEP within each cohort. To further investigate this possibility, changes in optical pathlength caused by changes in absorption we used SPSS Statistics 25 to perform a hierarchical cluster due to the passage of ICG through the joint. To address this, we analysis on the experimental group’s BF values from the first estimated the change in optical pathlength for five datasets and day of the baseline period (B1) to day 15 postinduction of dis- found that, on average, pathlength varied by 0.5% over a 120-s ease. Since no outliers were present in the dataset, Ward’s measurement period. When these differences were propagated method was used as the cluster method and squared Euclidean forward in the analysis, they only resulted in a 1% change distance as the similarity metric; the elbow method was sub- in recovered BF values. Since these changes are negligible, for sequently used to identify the optimal number of clusters.47 computational efficiency, p was estimated using the mean path- The analysis revealed that four distinct clusters were present length measured prior to ICG injection. Another limitation of in the data: one for each cohort in the experimental group. DCE TR-NIRS is that it requires the use of a contrast agent To confirm the stability of the solution, we randomized the order (e.g., ICG), which may limit clinical translation. However, it has of the data and reran the analysis three times. In addition, we been previously shown that changes in tissue oxy- and deoxy- also performed k-means clustering on the original dataset and hemoglobin concentrations in muscle following venous occlu- assumed that four clusters were present in the data. For all analy- sion can be used to quantify BF in lieu of a contrast agent.53 ses, it was determined that the same four clusters were present in While the ability of the latter method to quantify joint perfusion the data, with each cluster corresponding to our original cohort has yet to be tested, it offers an interesting alternative that could assignments. This provides further evidence that a main source be readily implemented with TR-NIRS instrumentation and of inconsistency in the study is that animals were divided into should be an area of future research. small cohorts whose arthritis was induced using four different Despite the advantages of the DCE TR-NIRS technique, the Mycobacterium butyricum solutions. Though every effort was small sample size of this exploratory work, issues with animal taken to ensure that all solutions were prepared using the same survivorship, and sources of error linked to disease hetero- protocol, it is possible that slight variations in injected solution geneity among the experimental group limit this study’s conclu- volume or preparation (e.g., mixing of bacteria with Freund’s sions regarding the link between BF, arthritis induction, and adjuvant) contributed to varied disease induction among the DMARD treatment response in the AIA model. Overall, we experimental group. found the use of the AIA model in a continuous, longitudinal As shown in this study, temporal variability in intersubject study design challenging due to the severe and rapid onset of disease and treatment response made it difficult to interpret the the disease, which often occurred in the 5-day periods between link between changes in BF and disease severity. Cases like this subsequent BF measurements. We also conducted a subset of benefit particularly from the ability of DCE TR-NIRS to reliably experiments to investigate mycobacterium dose and disease quantify tissue perfusion in absolute units, which makes it pos- severity but could not determine a reliable dose for inducing sible to compare a single subject’s values over a long range of slower and milder arthritic progression. Considering the mis- timepoints to uncover potential physiological relationships. It is match between this relatively rapid disease progression and the important to note that other optical techniques, such as diffuse relatively low overarching temporal resolution of our data (i.e., correlation spectroscopy (DCS), can noninvasively monitor one measurement every 5 days), our time courses may have

Journal of Biomedical Optics 015003-8 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

missed otherwise important variations in BF and disease activ- 4. J. A. Singh et al., “2015 American College of Rheumatology guideline ity. While the temporal frequency of our BF measurements was for the treatment of rheumatoid arthritis,” Arthritis Care Res. 68(1), 1–25 (2015). part of the original study design, which was formulated based “ 35 5. S. Monti et al., Rheumatoid arthritis treatment: the earlier the better to on results from our previous study, it was ultimately a limita- prevent joint damage,” RMD Open 1(Suppl. 1), e000057 (2015). tion of the study presented here and should be amended in future 6. Y. P. Goekoop-Ruiterman et al., “Clinical and radiographic outcomes of work. The authors also acknowledge the potentially con- four different treatment strategies in patients with early rheumatoid founding effect of anesthesia induced by isoflurane on BF in arthritis (the best study): a randomized, controlled trial,” Arthritis rodents.54,55 Though anesthetic concentrations were tightly Rheum. 52(11), 3381–3390 (2005). “ managed to consistent levels during experiments, future work 7. L. W. Moreland et al., A randomized comparative effectiveness study of oral triple therapy versus etanercept plus methotrexate in early, may benefit from investigating whether varying isoflurane con- aggressive rheumatoid arthritis,” Arthritis Rheumatol. 64(9), 2824– centrations or using other anesthetics such as ketamine has an 2835 (2012). effect on joint perfusion. Future work would also benefit from 8. K. P. Machold et al., “Very recent onset rheumatoid arthritis: clinical and exploring longitudinal BF changes in milder models of RA with serological patient characteristics associated with radiographic progres- slower symptom onset, such as pristane-induced arthritis,56 to sion over the first years of disease,” Rheumatology 46(2), 342–349 (2006). allow a more robust investigation of the relationship between “ DMARD treatment and joint BF. 9. P. Daul and J. Grisanti, Monitoring response to therapy in rheumatoid arthritis: perspectives from the clinic,” Bull. NYU Hosp. Joint Dis. 67(2), 236–242 (2009). 5 Conclusion 10. T. Uhlig, T. K. Kvien, and T. Pincus, “Test-retest reliability of disease activity core set measures and indices in rheumatoid arthritis,” Ann. In this work, we used our quantitative joint perfusion technique Rheum. Dis. 68(6), 972–975 (2009). (DCE TR-NIRS) to monitor joint BF in a longitudinal rat model 11. U. Døhn et al., “Detection of bone erosions in rheumatoid arthritis wrist of RA. Joint BF was measured in each animal’s left and right joints with magnetic resonance imaging, computed tomography and ankles before treatment (pretreatment phase) and after treatment radiography,” Arthritis Res. Ther. 10(1), R25 (2008). with the DMARD etanercept (treatment phase) in an experimen- 12. M. C. Micu et al., “Inter-observer reliability of ultrasound detection of tal group of eight rats with AIA. Four additional rats served as tendon abnormalities at the wrist and ankle in patients with rheumatoid arthritis,” Rheumatology 50(6), 1120–1124 (2011). controls throughout the study. Time and group had separate sig- 13. E. L. Rowbotham and A. J. Grainger, “Rheumatoid arthritis: ultrasound nificant effects on joint BF; however, there was no significant versus MRI,” Am. J. Roentgenol. 197(3), 541–546 (2011). interaction between time and group despite a large difference in 14. S. Torp-Pedersen et al., “Power and color Doppler ultrasound settings average BF values on day 5. Measurement side did not have a for inflammatory flow: impact on scoring of disease activity in patients significant effect on joint BF. Statistical analysis of the treatment with rheumatoid arthritis,” Arthritis Rheumatol. 67(2), 386–395 (2015). phase was limited since 50% of the animals reached HEPs 15. M. A. Akhavani et al., “Hypoxia upregulates angiogenesis and synovial ” before starting treatment. Comparison of individual animal time cell migration in rheumatoid arthritis, Arthritis Res. Ther. 11(3), R64 (2009). courses between the experimental and control group revealed no 16. S. Konisti, S. Kiriakidis, and E. M. Paleolog, “Hypoxia –a key regulator consistent trend in treatment response; effects may have been of angiogenesis and inflammation in rheumatoid arthritis,” Nat. Rev. masked by heterogeneous disease response to AIA in the exper- Rheumatol. 8(3), 153–162 (2012). imental group. Future work will focus on exploring longitudinal 17. P. C. Taylor and B. Sivakumar, “Hypoxia and angiogenesis in rheuma- BF changes in milder models of RA with slower symptom onset toid arthritis,” Curr. Opin. Rheumatol. 17(3), 293–298 (2005). to allow a more robust investigation of the relationship between 18. B. Sivakumar et al., “Synovial hypoxia as a cause of tendon rupture in ” – DMARD treatment and joint BF. rheumatoid arthritis, J. Hand Surg. 33(1), 49 58 (2008). 19. D. Bereczki et al., “Hypoxia increases velocity of blood flow through parenchymal microvascular systems in rat brain,” J. Cereb. Blood Flow Disclosures Metab. 13(3), 475–486 (1993). 20. C. A. Hitchon and H. S. El-Gabalawy, “Oxidation in rheumatoid arthri- The authors have no relevant conflicts of interest to disclose. tis,” Arthritis Res. Ther. 6(6), 265–278 (2004). 21. J. R. Rajian et al., “Characterization and treatment monitoring of inflam- Acknowledgments matory arthritis by photoacoustic imaging: a study on adjuvant-induced arthritis rat model,” Biomed. Opt. Express 4(6), 900–908 (2013). The authors would like to acknowledge the significant technical 22. G. Xu et al., “Photoacoustic and ultrasound dual-modality imaging for support of Lynn Keenliside throughout this project. This work inflammatory arthritis,” Proc. SPIE 8926, 892643 (2014). was supported by a Catalyst Grant from Western University’s 23. P. J. van den Berg et al., “Feasibility of photoacoustic/ultrasound im- Bone and Joint Institute, a Lawson Health Research Institute aging of synovitis in finger joints using a point-of-care system,” Internal Research Fund (IRF), an NSERC Discovery Grant Photoacoustics 8,8–14 (2017). “ (06337-2017 RGPIN), NSERC CGS-M, and the Collaborative 24. J. Jo et al., A functional study of human inflammatory arthritis using photoacoustic imaging,” Sci. Rep. 7(1), 1–9 (2017). Program in Musculoskeletal Health Research at Western 25. A. J. L. Meier et al., “Potential of optical spectral transmission measure- University. A preliminary version of the work presented here ments for joint inflammation measurements in rheumatoid arthritis 57 was published in SPIE Proceedings. patients,” J. Biomed. Opt. 17(08), 081420 (2012). 26. S. Ioussoufovitch et al., “Quantification of joint inflammation in rheumatoid arthritis by time-resolved diffuse optical spectroscopy and tracer References kinetic modeling,” Proc. SPIE 9319, 93191D (2015). 27. M. van Onna et al., “Assessment of disease activity in patients with 1. D. A. Walsh and D. F. McWilliams, “Pain in rheumatoid arthritis,” Curr. rheumatoid arthritis using optical spectral transmission measurements: Pain Headache Rep. 16(6), 509–517 (2012). a non-invasive imaging technique,” Ann. Rheum. Dis. 75(3), 511–518 2. F. Matcham et al., “The impact of rheumatoid arthritis on quality of life (2016). assessed using the SF-36: a systematic review and meta-analysis,” 28. V. Schafer et al., “Quantitative assessment of synovitis in patients with Semin. Arthritis Rheum. 44, 123–130 (2014). rheumatoid arthritis using fluorescence optical imaging,” Arthritis Res. 3. L. C. Pollard et al., “Fatigue in rheumatoid arthritis reflects pain, not Ther. 15(5), R124 (2013). disease activity,” Rheumatology 45(7), 885–889 (2006).

Journal of Biomedical Optics 015003-9 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use Ioussoufovitch et al.: Quantification of joint blood flow by dynamic contrast-enhanced near-infrared spectroscopy. . .

29. S.-Y. Kawashiri et al., “Indocyanine green (ICG)-enhanced fluorescence 45. M. Hegen et al., “Utility of animal models for identification of potential optical imaging (FOI) in patients with active rheumatoid arthritis; a therapeutics for rheumatoid arthritis,” Ann. Rheum. Dis. 67(11), 1505– comparative study with ultrasound and association with biomarkers,” 1515 (2008). Arthritis Rheumatol. 67, 267–269 (2015). 46. P. Totoson et al., “Etanercept improves endothelial function via pleio- 30. R. Meier et al., “Synovitis in patients with early inflammatory arthritis tropic effects in rat adjuvant-induced arthritis,” Rheumatology 55(7), monitored with quantitative analysis of dynamic contrast-enhanced 1308–1317 (2016). optical imaging and MR imaging,” Radiology 270(1), 176–185 (2014). 47. M. Sarstedt and E. Mooi, “Cluster analysis,” Chapter 9 in A Concise 31. J. M. Lasker et al., “Dynamic optical imaging of vascular and metabolic Guide to Market Research, M. Sarstedt and E. Mooi, Eds., 2nd ed., reactivity in rheumatoid joints,” J. Biomed. Opt. 12(5), 052001 (2007). pp. 273–324, Springer-Verlag Berlin Heidelberg, Heidelberg, Berlin 32. A. H. Hielscher et al., “Frequency-domain optical tomographic imaging (2014). of arthritic finger joints,” IEEE Trans. Med. Imaging 30(10), 1725–1736 48. M. Diop et al., “Calibration of diffuse correlation spectroscopy with a (2011). time-resolved near-infrared technique to yield absolute cerebral blood 33. L. D. Montejo et al., “Computer-aided diagnosis of rheumatoid arthritis flow measurements,” Biomed. Opt. Express 2(7), 2068–2082 (2011). with optical tomography, part 2: image classification,” J. Biomed. Opt. 49. M. Diop et al., “Development of a combined broadband near-infrared 18(7), 076002 (2013). and diffusion correlation system for monitoring cerebral blood flow and 34. H. Y. Wu et al., “Development of a multi-wavelength diffuse optical oxidative metabolism in preterm infants,” Biomed. Opt. Express 6(10), tomography system for early diagnosis of rheumatoid arthritis: simula- 3907 (2015). tion, phantoms and healthy human studies,” Biomed. Opt. Express 50. K. Verdecchia et al., “Assessment of the best flow model to characterize 7(11), 4769 (2016). diffuse correlation spectroscopy data acquired directly on the brain,” 35. A. Rajaram et al., “Joint blood flow is more sensitive to inflammatory Biomed. Opt. Express 6(11), 4288 (2015). arthritis than oxyhemoglobin, deoxyhemoglobin, and oxygen satura- 51. K. Verdecchia et al., “Quantifying the cerebral metabolic rate of oxygen tion,” Biomed. Opt. Express 7(10), 3843 (2016). by combining diffuse correlation spectroscopy and time-resolved near- 36. A. Bendele, “Animal models of rheumatoid arthritis,” J. infrared spectroscopy,” J. Biomed. Opt. 18(2), 027007 (2013). Musculoskeletal Neuronal Interact. 1(4), 377–385 (2001). 52. S. A. Carp et al., “Combined multi-distance frequency domain and dif- 37. W. van Eden, J. P. Wagenaar-Hilbers, and M. H. Wauben, “Adjuvant fuse correlation spectroscopy system with simultaneous data acquisition arthritis in the rat,” Curr. Protoc. Immunol. 19(1) (1996). and real-time analysis,” Biomed. Opt. Express 8(9), 3993 (2017). 38. P. Di Ninni, F. Martelli, and G. Zaccanti, “The use of India ink in tissue- 53. K. Gurley, Y. Shang, and G. Yu, “Noninvasive optical quantification of simulating phantoms,” Opt. Express 18(26), 26854–26865 (2010). absolute blood flow, blood oxygenation, and oxygen consumption rate 39. L. Spinelli et al., “Determination of reference values for optical proper- in exercising skeletal muscle,” J. Biomed. Opt. 17(7), 0750101 (2012). ties of liquid phantoms based on Intralipid and India ink,” Biomed. Opt. 54. W. C. Seyde and D. E. Longnecker, “Anesthetic influences on regional Express 5(7), 2037 (2014). hemodynamics in normal and hemorrhaged rats,” Anesthesiology 61(6), 40. S. A. Prahl and S. L. Jacques, “Optical properties spectra,” http://omlc 686–698 (1984). .ogi.edu/spectra/. 55. F. Kober et al., “Myocardial blood flow mapping in mice using “ 41. A. Cenic et al., Dynamic CT measurement of cerebral blood flow: a high-resolution spin labeling magnetic resonance imaging: influence ” – validation study, Am. J. Neuroradiol. 20(1), 63 73 (1999). of ketamine/xylazine and isoflurane anesthesia,” Magn. Reson. Med. “ 42. M. Diop et al., Time-resolved near-infrared technique for bedside mon- 53(3), 601–606 (2005). ” itoring of absolute cerebral blood flow, Proc. SPIE 7555, 75550Z 56. J. Tuncel et al., “Animal models of rheumatoid arthritis (I): pristane- (2010). induced arthritis in the rat,” PLoS One 11(5), e0155936 (2016). “ 43. M. Diop et al., Comparison of time-resolved and continuous-wave 57. S. Ioussoufovitch et al., “Quantifying joint blood flow in a rat model of ” near-infrared techniques for measuring cerebral blood flow in piglets, rheumatoid arthritis with dynamic contrast-enhanced near-infrared J. Biomed. Opt. 15(5), 057004 (2010). spectroscopy,” Proc. SPIE 10874, 108741S (2019). 44. F. Faul et al., “G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences,” Behav. Res. Biographies of the authors are not available. Methods 39(2), 175–191 (2007).

Journal of Biomedical Optics 015003-10 January 2020 • Vol. 25(1)

Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 09 Jul 2020 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use